Jiekun

Coding @VictoriaMetrics.
Learning Observability.
I contribute bugs to VictoriaMetrics.

[Jiekun @ReadTheDuck]

上个月随公司 #VictoriaMetrics 在 #KubeCon 上分享了如何在 #OpenTelemetry 中实践 Retroactive Sampling（回溯采样）。本着能写就不要浪费的原则，把演讲内容整理成了博客。

相比尾部采样，它能节约 60% 内存，考虑到现在内（金）存（子）的天价，应该或多或少有点意义 :)

https://t.co/W8zcsw1c1V

[Jiekun @ReadTheDuck]

Jaeger now has native support for ClickHouse, starting from v2.18. It's good to compare the performance with VictoriaTraces.

https://t.co/CvJsViLIyK

[Jiekun @ReadTheDuck]

@cloud728838 AI 只方便了发布内容的人，但是消耗了需要阅读和解决问题的人的时间，本来一个 5 秒钟看完就知道是什么��情的现在需要看包含了大量废话的“分析”。

不过可能是我没说清楚场景，主要在一些对话、帖子评论里，完全不应该让 AI 写作文。想象一下在飞书、微信问对方发生什么问题，然后回复一篇 800 字。

[Jiekun @ReadTheDuck]

对 AI 生成的分析文档和建议文档感到很恶心，一个一句话能讲完的事情写成了：
✅问题背景
✅原因分析
✅代码比对
✅修复建议

然后用的人还乐此不疲每个问题都按这样的格式来回复🤮现在看到这个绿色✅对钩就有种想���接 reject 的冲动 :(

[Darryl Ruggles @RDarrylR]

I like seeing real-world comparisons of various tools we use. Most log stack comparisons read like vendor pitches, so it's nice to see one grounded in actual production workloads. This one runs Loki and VictoriaLogs side by side on 500 GB of logs over 7 days, on identical hardware.

There are a handful of tools in this space, and Loki and VictoriaLogs are two of the more popular ones. The test results here are very interesting and I would like to try a similar comparison in my own environment. Harshit Luthra has put together a balanced comparison and outlines when each option is the right call, especially for label-heavy workloads and Grafana-native shops. Check it out!

https://t.co/W5EOFEgzZc

#VictoriaLogs #Loki #Grafana

[Jiekun @ReadTheDuck]

办信用卡的时候问我工作信息，答“软件工程师”。然后就紧接着这个问题，他问我工作用什么语言。我说：“Go 语言，Gee-Ooo，Go 语言。”

对方沉默了两秒然后说：“好的，这边记录好了”。

感谢对 DEI（Diversity, Equity, and Inclusion）的支持。🤔

[Jiekun @ReadTheDuck]

很早就想录一期正经的 Q&A 采访。

3月跟团队去参加 KubeCon 前，在公众号上向中国用户征集了 #VictoriaMetrics 相关问题现场解答。

划重点：远程办公还招人吗？VM何时支持对象存���？与 ClickHouse 竞争有没有压力？降采样等付费功能下放吗？

期待上海 KubeCon 线下交流！

https://t.co/gpk2hlzPFJ

[OpenObservability @OpenObserv]

📢 April episode is out! 📢
#VictoriaMetrics: From Monitoring to Observability

With @horovits hosting @hagen1778, Co-Founder and Engineering Manager @VictoriaMetrics

Listen on your favorite platform:
🎧 https://t.co/ctaLSoiMvA
📺 https://t.co/HtTSUTYVXe

[Jiekun @ReadTheDuck]

Thanks for sharing. We really enjoyed your practical experience and story!

[NAVER D2 @naver_d2]

<12.5억 개의 시계열과 555조 개의 데이터포인트를 다루는 VictoriaMetrics 아키텍처와 무중단 장비 전환>
네이버 검색은 VictoriaMetrics클러스터를 5년간 운영하며 메모리 한계를 어떻게 해결하고 180�� 장비 교체를 어떻게 서비스 중단 없이 마칠 수 있었는지 공유합니다
https://t.co/xVuKmDbuk0

[Jiekun @ReadTheDuck]

Our talk has the highest views among all #OpenTelemetry and Observability Day sessions at #KubeCon Europe 2026, nearly 2x as the 2nd place.

Thanks again to the team for putting me on stage.

目前我们的演讲在本次 KubeCon 的可观测性相关演讲中观看数最高，再次感谢团队将我带到舞台！

[VictoriaMetrics @VictoriaMetrics]

Last month, @readtheduck gave a talk on retroactive sampling at #KubeCon Europe 2026.
In this blog post, we explain how retroactive sampling reduces outbound traffic, CPU, and memory usage in the data collection pipeline significantly compared to tail sampling in #OpenTelemetry.
If you're working with observability at scale, this is worth a closer look.
👉 Learn more https://t.co/eOBtenoLLB

[Munshipremchand @MunshiPremChnd]

Airbnb migrates its high-volume metrics pipeline to OpenTelemetry, moving from StatsD and Veneur to OTLP, the OpenTelemetry Collector, and VictoriaMetrics vmagent. The new stack ingests 100M+ samples per second in production. Read the full story here: https://t.co/6dJsArLDdE

[Jiekun @ReadTheDuck]

The recording of our talk about Retroactive Sampling with #OpenTelemetry is available on Youtube now!

Slides here:
https://t.co/mlOS6aRCep

https://t.co/NxOIk8A1JL

[Jiekun @ReadTheDuck]

Few people print photos now. No cam film, no darkroom. If someone does it, there must be sth meaningful :) I didn't keep any team photos when working on-site as everyone was always sitting together. Remote work is different.
现在很少人还印照片了，如果有，那肯定是有什么值得记住的

[Jiekun @ReadTheDuck]

@strrlthedev 金价内存让这个事情变得更有趣了:) 开二手电脑店，是当代程序员唯一出路。

[Jiekun @ReadTheDuck]

上个月随公司 #VictoriaMetrics 在 #KubeCon 上分享了如何在 #OpenTelemetry 中实践 Retroactive Sampling（回溯采样）。本着能写就不要浪费的原则，把演讲内容整理成了博客。

相比尾部采样，它能节约 60% 内存，考虑到现在内（金）存（子）的天价，应该或多或少有点意义 :)

https://t.co/W8zcsw1c1V

[Jiekun @ReadTheDuck]

@foshuoai @hagen1778 没有只存储 status code 这一说吧，每个 span 都是完整的。精简的只是采样过程的数据，一旦确认被采样了，本地 Agent 需要提供完整的原始 span。希望我没理解错问题

[Jiekun @ReadTheDuck]

12点半从房间关电脑里走出来，有种小学的时候在网吧偷偷打游戏然后又下机的感觉😂🤣不是，可是上班跟上网明明不一样啊😇

[Jiekun @ReadTheDuck]

Good day @KubeCon_ Amsterdam, many thanks to everyone who’s joined our session retroactive sampling in #OpenTelemetry yesterday, and also big thanks to my co-speaker @hagen1778

[Jiekun @ReadTheDuck]

@valyala Thank you now someone will have the motivation to steal it, and rename it as vtagent.

[Jiekun @ReadTheDuck]

@xu_paco 今年还来 AMS 玩吗

[Roman Khavronenko @hagen1778]

One of the most thorough #Logs shippers benchmarks I've seen. Covering Vector, Promtail, Grafana Alloy, Fluent Bit, OpenTelemetry Collector, and more.

A benchmark suite, steps to reproduce, Grafana snapshots - all is there.

Well worth a read: https://t.co/ap5Dmv8PNj